From 0d6526f82c3cdefcca47f73f5fc08dc6f335eac6 Mon Sep 17 00:00:00 2001 From: Tim Chen Date: Mon, 21 Sep 2026 17:37:22 -0700 Subject: sched/cache: Keep nr_pref_llc_running in the runnable domain, to fix LLC mis-scheduling bug alb_break_llc() decides whether to break LLC preference during active load balance. It does so by testing that every runnable fair task on the source rq prefers its LLC: env->src_rq->nr_pref_llc_running == env->src_rq->cfs.h_nr_runnable But the two counters cover different sets. nr_pref_llc_running is updated in account_llc_enqueue()/account_llc_dequeue(), next to cfs_rq->nr_queued, so it follows queued tasks. h_nr_runnable is updated in set_delayed()/ clear_delayed() and drops delay-dequeued tasks. So under DELAY_DEQUEUE, a preferring task that goes to sleep stays counted in nr_pref_llc_running while h_nr_runnable falls. The equality then breaks, alb_break_llc() returns false, and active balance is free to pull a task off its preferred LLC. Active balance only moves runnable tasks, and this is the only LLC check it consults: once the stopper runs, LBF_ACTIVE_LB skips the per-task test in can_migrate_task(). The runnable set is the one we want. Fix it on the counter side. A task should be counted in nr_pref_llc_running exactly while it is both queued on its preferred LLC (pref_llc_queued) and runnable (!sched_delayed). Define that membership once in task_pref_llc_runnable(), and adjust the counter only through pref_llc_running_inc()/pref_llc_running_dec() from the four sites that change either input: account_llc_enqueue(), account_llc_dequeue(), set_delayed() and clear_delayed(). Gating every update on the same predicate keeps the delay, wake and dequeue paths from double-counting or underflowing; see the comments at those sites for the ordering. nr_llc_running and sd->llc_counts are not touched and stay on queued semantics. Fixes: 714059f79ff0 ("sched/cache: Handle moving single tasks to/from their preferred LLC") Closes: https://lore.kernel.org/lkml/20260827135000.735138-1-zhanxusheng@xiaomi.com/ Reported-by: Zhan Xusheng Suggested-by: Chen Yu Signed-off-by: Tim Chen Signed-off-by: Peter Zijlstra (Intel) Signed-off-by: Ingo Molnar Reviewed-by: Kayra Cizmeci Cc: # v7.2.x Link: https://patch.msgid.link/06af61afedac32e6477f57feb4d658f6c411c3af.1790035273.git.tim.c.chen@linux.intel.com --- arch/arm/mach-versatile/hotplug.c | 102 ++++++++++++++++++++++++++++++++++++++ 1 file changed, 102 insertions(+) create mode 100644 arch/arm/mach-versatile/hotplug.c (limited to 'arch/arm/mach-versatile/hotplug.c') diff --git a/arch/arm/mach-versatile/hotplug.c b/arch/arm/mach-versatile/hotplug.c new file mode 100644 index 000000000..5a1521755 --- /dev/null +++ b/arch/arm/mach-versatile/hotplug.c @@ -0,0 +1,102 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * Copyright (C) 2002 ARM Ltd. + * All Rights Reserved + * + * This hotplug implementation is _specific_ to the situation found on + * ARM development platforms where there is _no_ possibility of actually + * taking a CPU offline, resetting it, or otherwise. Real platforms must + * NOT copy this code. + */ +#include +#include +#include + +#include +#include + +#include "platsmp.h" + +static inline void versatile_immitation_enter_lowpower(unsigned int actrl_mask) +{ + unsigned int v; + + asm volatile( + "mcr p15, 0, %1, c7, c5, 0\n" + " mcr p15, 0, %1, c7, c10, 4\n" + /* + * Turn off coherency + */ + " mrc p15, 0, %0, c1, c0, 1\n" + " bic %0, %0, %3\n" + " mcr p15, 0, %0, c1, c0, 1\n" + " mrc p15, 0, %0, c1, c0, 0\n" + " bic %0, %0, %2\n" + " mcr p15, 0, %0, c1, c0, 0\n" + : "=&r" (v) + : "r" (0), "Ir" (CR_C), "Ir" (actrl_mask) + : "cc"); +} + +static inline void versatile_immitation_leave_lowpower(unsigned int actrl_mask) +{ + unsigned int v; + + asm volatile( + "mrc p15, 0, %0, c1, c0, 0\n" + " orr %0, %0, %1\n" + " mcr p15, 0, %0, c1, c0, 0\n" + " mrc p15, 0, %0, c1, c0, 1\n" + " orr %0, %0, %2\n" + " mcr p15, 0, %0, c1, c0, 1\n" + : "=&r" (v) + : "Ir" (CR_C), "Ir" (actrl_mask) + : "cc"); +} + +static inline void versatile_immitation_do_lowpower(unsigned int cpu, int *spurious) +{ + /* + * there is no power-control hardware on this platform, so all + * we can do is put the core into WFI; this is safe as the calling + * code will have already disabled interrupts. + * + * This code should not be used outside Versatile platforms. + */ + for (;;) { + wfi(); + + if (versatile_cpu_release == cpu_logical_map(cpu)) { + /* + * OK, proper wakeup, we're done + */ + break; + } + + /* + * Getting here, means that we have come out of WFI without + * having been woken up - this shouldn't happen + * + * Just note it happening - when we're woken, we can report + * its occurrence. + */ + (*spurious)++; + } +} + +/* + * platform-specific code to shutdown a CPU. + * This code supports immitation-style CPU hotplug for Versatile/Realview/ + * Versatile Express platforms that are unable to do real CPU hotplug. + */ +void versatile_immitation_cpu_die(unsigned int cpu, unsigned int actrl_mask) +{ + int spurious = 0; + + versatile_immitation_enter_lowpower(actrl_mask); + versatile_immitation_do_lowpower(cpu, &spurious); + versatile_immitation_leave_lowpower(actrl_mask); + + if (spurious) + pr_warn("CPU%u: %u spurious wakeup calls\n", cpu, spurious); +} -- cgit v1.3.1